Theory of Mind, Computational Tractability, and Mind Shaping

Philosophers and psychologists have traditionally understood “theory of mind” as a human capacity for understanding and predicting human behavior based on the attribution of unobservable mental states, like beliefs and desires. The classical model views the human “mind reader” as a kind of scientist, formulating hypotheses about the unobservable causes of the behavior of her fellows, and then testing them through observation (Gopnik & Wellman 1995). Many argue that the attribution of unobservable, theoretical mental states increases the power of human social cognition over mere sensitivity to patterns of observable behavior, of the kind that characterizes the social cognition of most, if not all non-human animals (Tomasello & Call 1997). In this paper, I review some familiar problems with this view and suggest a novel strategy for dealing with them. In section 2, I explain why the timely and accurate attribution of mental states appears to be a computationally intractable task. In sections 3 and 4, I consider two standard models of human cognitive architecture aimed at mitigating problems of computational tractability: modularity and fast and frugal heuristics, respectively. I argue that these are unlikely to help in the case of theory of mind. In the final section, I show how “mind shaping” (Mameli 2001; Zawidzki 2008) – roughly, the practice of socializing individuals in ways that make human populations more homogeneous – can mitigate some of the problems raised in the earlier sections.